Hydropower station GIL equipment insulation state detection and evaluation method based on industrial internet platform

Abstract

The invention discloses a hydropower station GIL equipment insulation state detection and evaluation method based on an industrial internet platform, and the method comprises the steps: collecting the characteristic parameters of voltage, current, temperature and the like through an online monitoring system, a live detection device or a power failure high-voltage test, and generating an original characteristic set; through photoelectric conversion, filtering and other processing, key features are extracted to generate a structured feature set, and the structured feature set is serialized and uploaded to a data center. A phase resolution graph and a dielectric spectrogram are calculated by utilizing technologies such as partial discharge analysis and a knowledge graph, fault diagnosis is carried out, and information such as fault positions and types is determined. And based on a diagnosis result, evaluating an insulation state risk through methods such as fault mode and influence analysis and fault tree analysis, quantifying defects, safety and economic risks, and outputting an alarm level and a processing suggestion. And if the risk is triggered, performing convolutional neural network analysis on the specific signal to update the fault position, and finally generating a diagnosis report according to the fault influence range, the diagnosis result and the alarm level, thereby realizing comprehensive monitoring and accurate evaluation of the insulation state of the GIL equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of hydropower equipment maintenance management, and particularly relates to a method for detecting and evaluating the insulation state of GIL equipment in a hydropower station based on an industrial Internet platform. Background Technique

[0002] In a hydropower station, gas-insulated transmission lines (GIL) are widely used as an efficient power transmission method due to their advantages such as compact structure, high transmission efficiency, and low maintenance costs. However, the insulation state of GIL equipment is directly related to the safe and stable operation of the power system. Monitoring and evaluating the insulation state is of great significance for preventing equipment failures, reducing unexpected power outages, and ensuring power supply. Traditional methods for detecting the insulation state of GIL equipment have limitations. For example, due to the fully enclosed structure of GIL equipment, it is difficult to detect internal state information. Once a failure occurs in the equipment, it will seriously affect the normal operation of the system. In addition, existing conventional technical means such as ultra-high frequency and ultrasonic waves are difficult to accurately detect internal defects of insulating components. Statistics show that GIS / GIL equipment failures are mainly related to insulation and mechanical defects generated during production, transportation, and operation, and mechanical defects are also important factors leading to insulation degradation.

[0003] In the context of the application of industrial Internet platforms, new requirements are put forward for the monitoring and evaluation methods of GIL equipment. Industrial Internet platforms can achieve remote monitoring, data analysis, and intelligent decision-making of equipment, providing new technical means for detecting and evaluating the insulation state of GIL equipment. However, how to effectively combine industrial Internet platforms with the detection and evaluation of the insulation state of GIL equipment to achieve precise monitoring and intelligent evaluation remains a technical challenge. Currently, research on methods for detecting and evaluating the insulation state of GIL equipment mainly focuses on partial discharge detection, ultrasonic guided wave detection, vibration signal monitoring, etc. These methods can reflect the insulation state of GIL equipment to a certain extent, but there are still problems such as limited detection range, insufficient accuracy, and inability to comprehensively evaluate the insulation state. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for detecting and evaluating the insulation state of GIL equipment in a hydropower station based on an industrial Internet platform, which can make full use of the advantages of industrial Internet platforms, combine multiple detection technologies, and achieve comprehensive and precise monitoring and evaluation of the insulation state of GIL equipment, so as to improve the operation safety and reliability of the power system.

[0005] To solve the above technical problem, the technical solution adopted by the present invention is: A method for detecting and evaluating the insulation state of GIL equipment in a hydropower station based on an industrial Internet platform, the steps are as follows: S1: Collect characteristic parameters of the insulation state of the GIL equipment through an on-line monitoring system, live detection device or outage high-voltage test, including voltage, current, temperature, vibration, UHF signal and ultrasonic signal, and generate an original feature set A, where A represents the initial data set containing all the collected signals; S2: Perform optoelectronic conversion, filtering and signal amplification processing on the original feature set A, combine time-domain analysis and frequency-domain analysis, extract vibration amplitude, UHF signal characteristic values and ultrasonic signal characteristic values, and generate a structured feature set B, where B represents the standardized data set containing the extracted features; S3: Serialize the structured feature set B into a JSON data packet in a predefined format, upload it to the substation data center through an encrypted transmission channel, and synchronize it to the company's big data center in real time to generate a stored feature set C, where C represents the serialized data set that has been stored; S4: Perform partial discharge analysis on the characteristic values in the stored feature set C, calculate the phase-resolved partial discharge pattern PRPD and the phase-resolved pulse sequence pattern PRPS, combine dielectric response analysis to generate a dielectric spectrogram, and generate a visualization atlas set D, where D represents the atlas set containing PRPD, PRPS and dielectric spectrogram; S5: Perform fault diagnosis on the visualization atlas set D based on the knowledge graph. The knowledge graph uses equipment components as nodes E = {e_1, e_2, …, e_m}, electrical connections and physical adjacencies as edges F, and combines PRPD and PRPS characteristic values to determine the fault location, type, gas chamber status and decomposition product concentration, and generate a diagnostic result set G, where G represents the structured set containing fault information; S6: Evaluate the insulation state risk based on the diagnostic results, analyze the defect trend, safety and economic risks, calculate the risk value through FMEA and FTA methods, and output the alarm level and treatment suggestions.

[0006] Preferably, the sub-steps of S1 are: S1.1: Collect voltage signals, current signals, temperature data, vibration characteristics, UHF signals and ultrasonic signals from the GIL equipment through on-line monitoring and live detection or outage tests, and generate an initial data set A; S1.2: Use preprocessing techniques to denoise and standardize the initial data set A to obtain a data set Q, where denoising is completed through wavelet transform and standardization uses the mean normalization method; S1.3: If the outliers of the voltage signal and current signal in the data set Q exceed the preset threshold, calculate the Pearson correlation coefficient between the voltage signal and the current signal through correlation analysis to determine whether there are potential defects in the insulation state; S1.4: According to the obtained dataset Q, extract the temperature data and vibration characteristics, and use the principal component analysis algorithm to reduce the dimension of the temperature data and vibration characteristics to obtain dataset R; S1.5: If the variance of the eigenvectors of the UHF signals and ultrasonic signals in dataset R is lower than the preset threshold, then classify the UHF signals and ultrasonic signals through cluster analysis to determine the category of the insulation state; S1.6: Obtain the Pearson correlation coefficient and category information, and use the support vector machine algorithm to perform classification training on dataset Q to obtain a prediction model for the insulation state; S1.7: Test dataset R through the prediction model to obtain the classification result of the insulation state of the GIL device.

[0007] Preferably, the sub-steps of S2 are: S2.1: Process the original feature set A through optoelectronic conversion, use an optoelectronic sensor to obtain initial signal data, and obtain the converted signal S; S2.2: Filter the converted signal S, use a low-pass filter to remove high-frequency noise, and obtain the filtered signal T; S2.3: Amplify the filtered signal T, use an operational amplifier to enhance the signal strength, and obtain the amplified signal U; S2.4: Perform time-domain analysis on the amplified signal U, calculate the time-series statistics of the signal, and obtain the vibration amplitude characteristic V; S2.5: Perform frequency-domain analysis on the amplified signal U, extract frequency components through fast Fourier transform, and obtain the UHF signal characteristic W and ultrasonic signal characteristic X; S2.6: If the integrity of the vibration amplitude characteristic V, UHF signal characteristic W, and ultrasonic signal characteristic X meets the preset threshold, then generate a structured feature set B through a feature fusion algorithm; S2.7: If not, return to the signal filtering step for reprocessing and judge the integrity of the feature set; S2.8: Standardize the structured feature set B, use the z-score normalization method, and obtain the standardized dataset Y.

[0008] Preferably, the sub-steps of S3 are: S3.1: Obtain the original data from the structured feature set, use a predefined serialization format to generate a JSON data packet, and obtain the serialized dataset; S3.2: Encrypt the JSON data packet through an encryption algorithm to generate an encrypted data packet and determine the transmission security; S3.3: If the integrity verification of the encrypted data packet passes, it is transmitted to the substation data center through the encrypted channel, and the transmission completion status is obtained; S3.4: Receive the encrypted data packet from the substation data center, decrypt it to generate the stored feature set, and obtain the stored data set; S3.5: If the stored feature set is consistent with the serialized data set, it is sent to the big data center through the synchronous transmission protocol, and the synchronization completion status is determined; S3.6: According to the stored feature set received by the big data center, use the data verification algorithm to judge the data integrity; S3.7: If the data integrity verification passes, store the feature set in the big data center and obtain the final storage status.

[0009] Preferably, the sub-steps of S4 are: S4.1: Obtain the feature values from the stored feature set C, calculate the phase-resolved features through partial discharge analysis, and obtain the initial data of the PRPD diagram and the PRPS diagram; S4.2: Process the initial data of the PRPD diagram using Fourier transform, extract the phase-resolved discharge features, and obtain the frequency-domain distribution of the discharge diagram; S4.3: Process the initial data of the PRPS diagram through time series analysis, determine the time distribution characteristics of the pulse sequence, and obtain the time-domain characteristics of the pulse sequence diagram; S4.4: Extract the dielectric response data from the feature values, use the dielectric spectroscopy analysis algorithm to generate the frequency-domain response distribution of the dielectric spectrogram; S4.5: If the frequency-domain distribution of the PRPD diagram matches the preset discharge feature threshold, combine the time-domain characteristics of the PRPS diagram to generate a joint feature set of the discharge mode; S4.6: Obtain the joint feature set and the frequency-domain response distribution of the dielectric spectrogram, use the data fusion algorithm to obtain the visualization atlas set D including the PRPD diagram, the PRPS diagram, and the dielectric spectrogram; S4.7: Through visualization processing, convert the atlas set D into a two-dimensional image output to obtain the final visualization result.

[0010] Preferably, the sub-steps of S5 are: S5.1: By parsing the visualization atlas set D, obtain the device component nodes E = {e_1, e_2,..., e_m} and the electrical connection and physical adjacency edges F from the knowledge graph, construct the topological relationship matrix Z, and obtain the connection features between the device components; S5.2: According to the topological relation matrix Z, fuse the PRPD eigenvalue, and use the k-nearest neighbor algorithm to calculate the abnormal probability P(e_i) of each device component. If P(e_i) is greater than the preset threshold T_1, determine that the component e_i is a potential fault location, and obtain the fault location set α; S5.3: For each component e_i in the fault location set α, extract the corresponding PRPS eigenvalue, classify the fault mode using the support vector machine algorithm, and determine whether the fault type is breakdown, discharge or overheating, to obtain the fault type set β; S5.4: Obtain the chamber state parameter S(e_i) of the component e_i from the fault location set α. If S(e_i) deviates from the preset threshold range T_2, determine that the chamber state is abnormal, and obtain the chamber state set γ; S5.5: According to the fault type set β and the chamber state set γ, analyze the decomposition product characteristics. If the fault type is discharge and the chamber state is abnormal, calculate the decomposition product concentration C using the spectral analysis technology, and obtain the decomposition product concentration set δ; S5.6: By fusing the fault location set α, the fault type set β, the chamber state set γ and the decomposition product concentration set δ, generate a structured fault diagnosis result set G, where G = {α, β, γ, δ}; S5.7: If there are multi-component fault associations in the fault diagnosis result set G, based on the electrical connection edge F of the knowledge graph, use the graph convolutional network to update the fault propagation path, and obtain the final diagnosis result set G.

[0011] Preferably, the sub-steps of S6 are: S6.1: According to the diagnosis result set G, perform a fault mode and effects analysis, identify each fault mode and its impact range on the system performance and component operation, and combine with the edge F of the knowledge graph to generate a component association impact set H, where H represents the mapping set of the fault mode and the affected components; S6.2: According to the component association impact set H, perform a fault tree analysis, define insulation failure as the top event, use the fault modes in H as the basic events, construct the causal propagation path, and generate a fault propagation graph I, where I represents the logical relationship set from component failure to system failure; S6.3: According to the diagnosis result set G and the fault propagation graph I, quantify the defect trend, safety risk and economic risk of the insulation state, calculate the comprehensive risk value J, where J represents the risk quantification value, generate an alarm level K and a handling suggestion L, where K represents the risk level and L represents the set of suggested measures; S6.4: If the alarm level K is higher than the preset threshold M, where M represents the risk trigger threshold, then perform convolutional neural network analysis on the UHF signals and ultrasonic signals in the structured feature set B. The convolutional kernel is based on a pre-trained model to generate the fault location classification probability N, where N represents the location verification probability, and update the fault location in the diagnostic result set G; S6.5: Based on the fault location in the updated diagnostic result set G, query the associated device components through the knowledge graph. The edge F is defined as the electrical dependency relationship to generate the fault impact range O, where O represents the set of components affected by the fault {e_i, e_j,...}; S6.6: According to the fault impact range O, the diagnostic result set G, and the alarm level K, generate a diagnostic report P based on the predefined alarm rules. P includes the fault type, location, impact range, and handling suggestions, and transmit P to the monitoring system to complete the insulation status monitoring.

[0012] Preferably, the specific method of S6.1 is: Obtain the fault mode data from the diagnostic result set, parse and extract the feature information to form a fault mode set; for this set, combine the system performance parameters and the component operation status, and use the impact analysis algorithm to determine the impact range of each fault mode on the system and components; with the help of the knowledge graph, obtain the component association relationship based on the edge F to generate a preliminary association impact set of the fault mode and components; if there are multi-path impacts in the preliminary set, use the depth-first search algorithm to traverse the edge F to judge the propagation path and obtain an optimized association impact set; construct a mapping relationship between the fault mode and the affected components according to the optimized set to form a component mapping set; extract the identification information of the affected components from the component mapping set, and combine the system performance change trend to judge the priority to obtain the final component association impact set H; verify the consistency of H and the diagnostic result set through data comparison, determine the accuracy of H and output the final mapping set.

[0013] Preferably, the specific method of S6.2 is: Obtain the component association impact set H, and extract the fault mode from it to form a basic event set ε; use the fault tree analysis method, set the insulation failure as the top event ζ, and construct a fault tree structure η based on ε; through Boolean algebra operations, analyze the causal propagation relationship between the basic events in η to generate a causal propagation path θ; calculate the logical relationship strength ι of each path according to θ to obtain a fault propagation map I; if the ι of a certain path in I exceeds the preset threshold, mark this path as a high-risk path κ; use κ to extract the key basic event λ to generate an optimized fault tree μ; finally, output the system failure logical relationship set ν of the complete causal chain from the key component failure to the insulation failure according to μ.

[0014] Preferably, the specific method of S6.3 is: Extract insulation status data from the diagnostic result set G, and use statistical analysis to obtain defect characteristics and change trends; combine the defect change trends and the fault propagation map I, and use graph analysis algorithms to calculate the defect propagation paths and clarify the distribution of potential fault points; if the distribution of potential fault points exceeds the preset threshold, obtain the safety risk value ξ through the safety risk model to complete the quantification of safety risks; use the economic loss assessment model, and combine the defect change trends to calculate the economic risk value ο to achieve the quantification of economic risks; adopt a weighted fusion method, and comprehensively combine the safety risk value ξ and the economic risk value ο to obtain the comprehensive risk value J, and determine the comprehensive risk quantification index; if the comprehensive risk value J exceeds the preset threshold, generate the alarm level K according to the risk grading rules; finally, according to the alarm level K, extract the corresponding set of processing measures L from the preset measure library to generate recommended measures.

[0015] The present invention can achieve the following beneficial effects: 1. By integrating various sensing technologies and the data processing capabilities of the industrial Internet platform, this method can achieve real-time and comprehensive monitoring of the insulation status of GIL equipment, improving the accuracy and efficiency of detection.

[0016] 2. Using the big data analysis function of the industrial Internet platform, this method can deeply analyze a large amount of collected data, so as to more accurately predict potential faults and insulation degradation trends.

[0017] 3. Through continuous monitoring and evaluation of GIL equipment, this method helps to detect and handle early faults in a timely manner, avoid large-scale equipment damage, and thus significantly reduce maintenance costs and repair expenses.

[0018] 4. This method can evaluate the insulation status of GIL equipment in real time, detect potential safety hazards in a timely manner, reduce unexpected power outages, and thus improve the reliability and stability of the entire power system.

[0019] 5. Combined with the intelligent decision support system of the industrial Internet platform, this method can automatically generate maintenance and repair suggestions to achieve intelligent management of GIL equipment.

[0020] 6. Through effective monitoring and timely maintenance, this method helps to extend the service life of GIL equipment, reduce the equipment replacement frequency, and lower the long-term operation cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The following further describes the present invention in conjunction with the drawings and embodiments: Figure 1 It is the flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The preferred solution is as Figure 1As shown in the figure, a method for detecting and evaluating the insulation status of GIL equipment in hydropower stations based on an industrial Internet platform is a method and strategy for detecting and evaluating the insulation status of GIL equipment in hydropower stations based on industrial Internet platform technology. This method can use relevant components, algorithms, data, etc. of the industrial Internet platform to achieve more efficient detection and evaluation of the insulation status of equipment. The structures included are the signal perception layer, the front-end processing layer, the signal transmission layer, the data processing layer, the diagnostic model layer, the status evaluation layer, etc. Signal perception layer. This layer is responsible for collecting characteristic parameters reflecting the insulation status of GIL equipment, including but not limited to equipment voltage and current, temperature and humidity, vibration, ultra-high frequency, ultrasonic wave, gas decomposition products, gas density, capacitance, dielectric spectrum, insulation resistance, DC resistance, etc. The collection means include on-line monitoring systems, live detection devices, out-of-service high-voltage tests, etc.; Front-end processing layer. The collected signals are preliminarily processed, including photoelectric conversion, filtering, amplification, time-domain analysis, frequency-domain analysis, etc., to form data such as vibration amplitude, ultra-high frequency, ultrasonic wave, etc. Signal transmission layer. The data processed by the front end is uploaded to the substation data center in accordance with the specified standard format, and then uploaded to the company's big data center through the substation data center. Data processing layer. The uploaded data is processed to form various spectrograms, such as PRPD, PRPS, FPRPD, original waveform, dielectric spectrogram, correlation comparison diagram, etc. The diagnostic model layer uses knowledge graph technology to diagnose defects in the formed spectrograms, including the location, type, gas chamber where the discharge source is located, gas chamber pressure change trend, discharge source development trend, gas chamber decomposition product concentration, etc. Status evaluation layer. Through information such as the defect location, type, trend, cause, safety risk, economic risk, etc. obtained from the diagnosis, a risk assessment of the insulation status is carried out, and treatment suggestions are output. The risk value of the insulation status of GIL equipment is calculated using FMEA, FTA, etc., the alarm level is output, and treatment suggestions for different levels are given. The present invention provides a specific method for detecting and evaluating the insulation status of GIL equipment in hydropower stations. Through the above technical solutions, a comprehensive detection and evaluation of the insulation status of GIL equipment is achieved, effectively improving the safety and stability of the power system. The specific implementation steps are as follows: Signal perception layer: S1: Through an on-line monitoring system, a live detection device or an out-of-service high-voltage test, collect characteristic parameters of the insulation status of GIL equipment, including voltage, current, temperature, vibration, ultra-high frequency signal and ultrasonic signal, and generate an original feature set A, where A represents the initial data set containing all the collected signals.

[0023] S1.1: Through on-line monitoring, live detection or out-of-service tests, collect voltage signals, current signals, temperature data, vibration characteristics, ultra-high frequency signals and ultrasonic signals from GIL equipment, and generate an initial data set A.

[0024] S1.2: Denoise and standardize the initial dataset A using preprocessing techniques to obtain dataset Q, where denoising is accomplished through wavelet transform and standardization adopts the mean normalization method.

[0025] S1.3: If the outliers of the voltage signal and current signal in dataset Q exceed the preset threshold, calculate the Pearson correlation coefficient between the voltage signal and current signal through correlation analysis to determine whether there are potential defects in the insulation state.

[0026] S1.4: According to the obtained dataset Q, extract temperature data and vibration characteristics, and use the principal component analysis algorithm to reduce the dimension of the temperature data and vibration characteristics to obtain dataset R.

[0027] S1.5: If the variance of the eigenvectors of the UHF signal and ultrasonic signal in dataset R is lower than the preset threshold, classify the UHF signal and ultrasonic signal through clustering analysis to determine the category of the insulation state.

[0028] S1.6: Obtain the Pearson correlation coefficient and category information, and use the support vector machine algorithm to classify and train dataset Q to obtain a prediction model for the insulation state.

[0029] S1.7: Test dataset R through the prediction model to obtain the classification result of the insulation state of the GIL device.

[0030] Front-end processing layer: S2: Perform optoelectronic conversion, filtering, and signal amplification on the original feature set A, combine time-domain analysis and frequency-domain analysis, extract vibration amplitude, UHF signal eigenvalue, and ultrasonic signal eigenvalue, and generate a structured feature set B, where B represents a standardized dataset containing the extracted features.

[0031] S2.1: Process the original feature set A through optoelectronic conversion, and use an optoelectronic sensor to obtain initial signal data to get the converted signal S.

[0032] S2.2: Filter the converted signal S, and use a low-pass filter to remove high-frequency noise to get the filtered signal T.

[0033] S2.3: Amplify the filtered signal T, and use an operational amplifier to enhance the signal strength to get the amplified signal U.

[0034] S2.4: Perform time-domain analysis on the amplified signal U, calculate the time-series statistics of the signal to get the vibration amplitude feature V.

[0035] S2.5: Perform frequency-domain analysis on the amplified signal U, extract frequency components through fast Fourier transform to get the UHF signal feature W and ultrasonic signal feature X.

[0036] S2.6: If the integrity of the vibration amplitude feature V, the UHF signal feature W, and the ultrasonic signal feature X meets the preset threshold, generate a structured feature set B through a feature fusion algorithm.

[0037] S2.7: If not, return to the signal filtering step for reprocessing to judge the integrity of the feature set.

[0038] S2.8: Perform normalization processing on the structured feature set B using the z - score normalization method to obtain a normalized data set Y.

[0039] Signal transmission layer: S3: Serialize the structured feature set B into a JSON data packet according to a predefined format, upload it to the substation data center through an encrypted transmission channel, and synchronize it to the company's big data center in real time to generate a stored feature set C, where C represents the stored serialized data set.

[0040] S3.1: Obtain the original data from the structured feature set, generate a JSON data packet using a predefined serialization format, and obtain a serialized data set.

[0041] S3.2: Encrypt the JSON data packet through an encryption algorithm to generate an encrypted data packet and determine the transmission security.

[0042] S3.3: If the integrity verification of the encrypted data packet passes, transmit it to the substation data center through an encrypted channel to obtain the transmission completion status.

[0043] S3.4: Receive the encrypted data packet from the substation data center, decrypt it to generate a stored feature set, and obtain a stored data set.

[0044] S3.5: If the stored feature set is consistent with the serialized data set, send it to the big data center through a synchronous transmission protocol to determine the synchronization completion status.

[0045] S3.6: According to the stored feature set received by the big data center, use a data verification algorithm to judge the data integrity.

[0046] S3.7: If the data integrity verification passes, store the feature set in the big data center to obtain the final storage status.

[0047] Data processing layer: S4: Perform partial discharge analysis on the feature values in the stored feature set C, calculate the phase - resolved partial discharge pattern PRPD and the phase - resolved pulse sequence pattern PRPS, generate a dielectric spectrogram by combining dielectric response analysis, and generate a visualization atlas set D, where D represents the atlas set containing PRPD, PRPS, and the dielectric spectrogram.

[0048] S4.1: Obtain the eigenvalue from the stored feature set C, calculate the phase-resolved feature through partial discharge analysis, and obtain the initial data of the PRPD diagram and the PRPS diagram.

[0049] S4.2: Process the initial data of the PRPD diagram using Fourier transform, extract the phase-resolved discharge feature, and obtain the frequency-domain distribution of the discharge diagram.

[0050] S4.3: Process the initial data of the PRPS diagram through time-series analysis, determine the time distribution characteristics of the pulse sequence, and obtain the time-domain feature of the pulse sequence diagram.

[0051] S4.4: Extract the dielectric response data from the eigenvalue, adopt the dielectric spectroscopy analysis algorithm, and generate the frequency-domain response distribution of the dielectric spectrogram.

[0052] S4.5: If the frequency-domain distribution of the PRPD diagram matches the preset discharge feature threshold, then combine the time-domain feature of the PRPS diagram to generate a joint feature set of the discharge mode.

[0053] S4.6: Obtain the joint feature set and the frequency-domain response distribution of the dielectric spectrogram, adopt the data fusion algorithm, and obtain the visualization atlas set D including the PRPD diagram, the PRPS diagram, and the dielectric spectrogram.

[0054] S4.7: Through visualization processing, convert the atlas set D into a two-dimensional image for output to obtain the final visualization result.

[0055] Diagnostic model layer: S5: Perform fault diagnosis on the visualization atlas set D based on the knowledge graph. The knowledge graph takes the equipment components as nodes E = {e_1, e_2, …, e_m}, takes the electrical connection and physical adjacency as edges F, combines the PRPD and PRPS eigenvalues, determines the fault location, type, chamber state, and decomposition product concentration, and generates a diagnostic result set G, where G represents a structured set containing fault information.

[0056] S5.1: By parsing the visualization atlas set D, obtain the equipment component nodes E = {e_1, e_2, …, e_m} and the electrical connection and physical adjacency edges F from the knowledge graph, construct a topological relation matrix Z, and obtain the connection features between equipment components.

[0057] S5.2: According to the topological relation matrix Z, fuse the PRPD eigenvalues, adopt the k-nearest neighbor algorithm to calculate the abnormal probability P(e_i) of each equipment component. If P(e_i) is greater than the preset threshold T_1, then determine the component e_i as a potential fault location to obtain the fault location set α.

[0058] S5.3: For each component \(e_i\) in the fault location set \(\alpha\), extract the corresponding PRPS eigenvalue, classify the fault mode using the support vector machine algorithm, and determine whether the fault type is breakdown, discharge, or overheating to obtain the fault type set \(\beta\).

[0059] S5.4: Obtain the chamber state parameter \(S(e_i)\) of the component \(e_i\) from the fault location set \(\alpha\). If \(S(e_i)\) deviates from the preset threshold range \(T_2\), then determine that the chamber state is abnormal to obtain the chamber state set \(\gamma\). S5.5: According to the fault type set \(\beta\) and the chamber state set \(\gamma\), analyze the decomposition product characteristics. If the fault type is discharge and the chamber state is abnormal, then calculate the decomposition product concentration \(C\) using the spectral analysis technique to obtain the decomposition product concentration set \(\delta\).

[0060] S5.6: By fusing the fault location set \(\alpha\), the fault type set \(\beta\), the chamber state set \(\gamma\), and the decomposition product concentration set \(\delta\), generate a structured fault diagnosis result set \(G\), where \(G = \{\alpha, \beta, \gamma, \delta\}\).

[0061] S5.7: If there are multi-component fault associations in the fault diagnosis result set \(G\), based on the electrical connection edge \(F\) of the knowledge graph, use the graph convolutional network to update the fault propagation path to obtain the final diagnosis result set \(G\).

[0062] State evaluation layer: S6: Evaluate the insulation state risk based on the diagnosis results, analyze the defect trend, safety, and economic risks, calculate the risk values through methods such as FMEA and FTA, and output the warning level and handling suggestions.

[0063] S6.1: According to the diagnosis result set \(G\), perform a failure mode and effects analysis (FMEA) to identify each failure mode and its scope of influence on system performance and component operation, and combine with the edge \(F\) of the knowledge graph to generate a component association impact set \(H\), where \(H\) represents the mapping set of the failure mode and the affected components.

[0064] S6.1.1: Obtain the failure mode data from the diagnosis result set, and parse and extract the characteristic information of each failure mode to obtain the failure mode set.

[0065] S6.1.2: For the failure mode set, use the impact analysis algorithm, combine with the system performance parameters and component operation status, and determine the scope of influence of each failure mode on system performance and component operation.

[0066] S6.1.3: Through the knowledge graph, obtain the component association relationship represented by the graph edge \(F\) to generate a preliminary association impact set of the failure mode and components.

[0067] S6.1.4: If there are multi-path effects in the preliminary associated impact set, the depth-first search algorithm is used to traverse the graph edges F to determine the propagation path of the failure mode, and the optimized associated impact set is obtained.

[0068] S6.1.5: According to the optimized associated impact set, construct the mapping relationship from the failure mode to the affected components, and generate the component mapping set.

[0069] S6.1.6: Extract the identification information of the affected components from the component mapping set, and combine it with the change trend of the system performance to judge the priority of the affected components, and obtain the final component associated impact set H.

[0070] S6.1.7: By verifying the consistency between the component associated impact set H and the diagnostic result set, using the data comparison method, determine the accuracy of H, and output the final mapping set.

[0071] S6.2: According to the component associated impact set H, perform fault tree analysis (FTA), define insulation failure as the top event, use the failure modes in H as the basic events, construct the causal propagation path, and generate the fault propagation graph I, where I represents the set of logical relationships from component failure to system failure.

[0072] S6.2.1: Obtain the component associated impact set H, extract the failure modes from it, and generate the basic event set ε. ε represents all possible initial fault points that may cause system failure.

[0073] S6.2.2: Through the fault tree analysis method, set insulation failure as the top event ζ, and construct the fault tree structure η based on ε. η represents the logical connection from the basic event to the top event.

[0074] S6.2.3: Use Boolean algebra operations to analyze the causal propagation relationship between the basic events in η, and generate the causal propagation path θ. θ represents the transfer order of the fault from the component to the system.

[0075] S6.2.4: According to θ, calculate the logical relationship strength ι of each path. ι represents the likelihood weight of the fault propagation, and obtain the fault propagation graph I.

[0076] S6.2.5: If the ι of a certain path in I exceeds the preset threshold, mark this path as a high-risk path κ. κ represents the key propagation chain of system failure.

[0077] S6.2.6: Extract the key basic events λ through κ. λ represents the failure mode that has the greatest impact on system failure, and generate the optimized fault tree μ.

[0078] S6.2.7: According to μ, output the set of logical relationships ν of system failure. ν represents the complete causal chain from the key component failure to insulation failure.

[0079] S6.3: According to the diagnostic result set G and the fault propagation map I, quantify the defect trend, safety risk, and economic risk of the insulation state, calculate the comprehensive risk value J, where J represents the risk quantification value, generate the alarm level K and the handling suggestions L, where K represents the risk level and L represents the set of suggested measures.

[0080] S6.3.1: Obtain the insulation state data from the diagnostic result set G, extract the defect characteristics using statistical analysis methods, and obtain the defect change trend.

[0081] S6.3.2: According to the defect change trend, combined with the fault propagation map I, use graph analysis algorithms to calculate the defect propagation path and determine the distribution of potential fault points.

[0082] S6.3.3: If the distribution of potential fault points exceeds the preset threshold, calculate the safety risk value ξ through the safety risk model to obtain the safety risk quantification result.

[0083] S6.3.4: Through the economic loss assessment model, combined with the defect change trend, calculate the economic risk value ο to obtain the economic risk quantification result.

[0084] S6.3.5: Use the weighted fusion method, combined with the safety risk value ξ and the economic risk value ο, to calculate the comprehensive risk value J and determine the comprehensive risk quantification index.

[0085] S6.3.6: If the comprehensive risk value J exceeds the preset threshold, generate the alarm level K according to the risk grading rules to obtain the risk level.

[0086] S6.3.7: According to the alarm level K, extract the corresponding set of handling measures L from the preset measure library to generate suggested measures.

[0087] S6.4: If the alarm level K is higher than the preset threshold M, where M represents the risk trigger threshold, perform convolutional neural network analysis on the UHF signals and ultrasonic signals in the structured feature set B. The convolutional kernel is based on a pre-trained model to generate the fault location classification probability N, where N represents the location verification probability, and update the fault location in the diagnostic result set G.

[0088] S6.4.1: If the alarm level K exceeds the risk threshold M, where M represents the preset risk trigger threshold, extract the UHF signals and ultrasonic signals from the structured feature set B to obtain the initial signal set π.

[0089] S6.4.2: Through preprocessing the initial signal set π, perform normalization operations on the UHF signals and ultrasonic signals using the standardization method to obtain the standardized signal set ρ.

[0090] S6.4.3: Perform convolutional neural network analysis on the normalized signal set ρ using a pre-trained convolutional kernel to generate a fault location classification probability N, where N represents the location verification probability, and obtain a fault location distribution σ.

[0091] S6.4.4: If the maximum probability value in the fault location distribution σ is higher than a preset threshold T_loc, where T_loc represents the location confirmation threshold, then extract the location corresponding to the highest probability from the fault location distribution σ, update the fault location in the diagnosis result set G, and obtain an updated diagnosis set τ.

[0092] S6.4.5: According to the updated diagnosis set τ, use a clustering algorithm to group the fault locations and generate a fault area set υ, where υ represents the spatial distribution of the faults.

[0093] S6.4.6: By performing feature mapping on the fault area set υ, use the principal component analysis method to extract key features and generate a feature vector φ, where φ represents the dimensionality-reduced representation of the fault area.

[0094] S6.4.7: If the similarity between the feature vector φ and the historical fault library H_lib is higher than a preset threshold T_sim, where T_sim represents the fault matching threshold, then obtain the corresponding fault pattern from the historical fault library H_lib, update the fault area set υ, and obtain the final fault distribution χ.

[0095] S6.5: Based on the fault locations in the updated diagnosis result set G, query associated device components through a knowledge graph. The edge F is defined as an electrical dependency relationship, and generate a fault impact range O, where O represents the set of components {e_i, e_j,...} affected by the fault.

[0096] S6.5.1: Obtain the fault location identifier f_id from the diagnosis result set G.

[0097] S6.5.2: If there is an associated node for f_id in the knowledge graph, then extract the electrical dependency edge F directly connected to f_id by querying the knowledge graph to obtain an initial associated component set ψ = {e_i, e_j,...}.

[0098] S6.5.3: According to the initial associated component set ψ, use a depth-first search algorithm to traverse the electrical dependency edges F related to each component e_i in ψ in the knowledge graph to obtain an indirectly associated component set ω = {e_k, e_l,...}.

[0099] S6.5.4: Generate a merged component set ξ_merge = {e_i, e_j, e_k, e_l,...} by merging the initial associated component set ψ and the indirectly associated component set ω.

[0100] S6.5.5: For each component e_m in the merged component set ξ_merge, if the weight w_m of the electrical dependency edge F of e_m is greater than the preset threshold, then determine e_m as a component affected by the fault, and obtain the candidate affected component set ζ_cand = {e_m, e_n,...}.

[0101] S6.5.6: From the candidate affected component set ζ_cand, obtain the identification and status information of each component e_n, and through set deduplication processing, generate the final set O of components affected by the fault = {e_i, e_j,...}.

[0102] S6.5.7: According to the set O of components affected by the fault, use the topological sorting algorithm to arrange the components in O in the dependency order, and obtain the ordered affected component sequence η_seq = {e_i -> e_j ->...}.

[0103] S6.5.8: Through the ordered affected component sequence η_seq, generate the structured data θ_struct of the fault influence range, including the identification, status and dependency order of each component, and determine the final output.

[0104] S6.6: According to the fault influence range O, the diagnostic result set G and the alarm level K, generate a diagnostic report P according to the predefined alarm rules, where P includes the fault type, location, influence range and handling suggestions, and transmit P to the monitoring system to complete the insulation status monitoring.

[0105] S6.6.1: Obtain the abnormal data area from the fault influence range O, and determine the preliminary fault characteristics by comparing with the predefined rules.

[0106] S6.6.2: According to the preliminary fault characteristics, extract the relevant fault modes from the diagnostic result set G, and use the classification algorithm to judge the fault type.

[0107] S6.6.3: If the fault type is confirmed, extract the location parameters from the abnormal data area to obtain the fault location.

[0108] S6.6.4: Through the fault type and the fault location, combined with the alarm level K, use rule matching to obtain the handling suggestions.

[0109] S6.6.5: Generate a diagnostic report P according to the handling suggestions, including the fault type, fault location, influence range and handling suggestions.

[0110] S6.6.6: Extract the structured data from the diagnostic report P and transmit it to the monitoring system through the data transmission protocol.

[0111] S6.6.7: Obtain the feedback from the monitoring system and use the status monitoring algorithm to update the insulation status.

[0112] The above embodiments are only the preferred technical solutions of the present invention and should not be regarded as limitations on the present invention. The protection scope of the present invention shall be the technical solutions recited in the claims, including the equivalent replacement solutions of the technical features in the technical solutions recited in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for detecting and evaluating the insulation status of a GIL device in a hydropower station based on an industrial Internet platform, characterized in that It includes the following steps: S1: Collect the characteristic parameters of the insulation state of the GIL equipment through an on-line monitoring system, live detection device or outage high-voltage test, including voltage, current, temperature, vibration, UHF signal and ultrasonic signal, and generate the original feature set A, where A represents the initial data set containing all the collected signals; S2: Perform optoelectronic conversion, filtering and signal amplification processing on the original feature set A, and combine time-domain analysis and frequency-domain analysis to extract the vibration amplitude, UHF signal characteristic value and ultrasonic signal characteristic value, and generate the structured feature set B, where B represents the standardized data set containing the extracted features; S3: Serialize the structured feature set B into a JSON data packet in a predefined format, upload it to the substation data center through an encrypted transmission channel, and synchronize it to the company's big data center in real time to generate the stored feature set C, where C represents the serialized data set that has been stored; S4: Perform partial discharge analysis on the characteristic values in the stored feature set C, calculate the phase-resolved partial discharge pattern PRPD and the phase-resolved pulse sequence pattern PRPS, and generate a dielectric spectrogram by combining dielectric response analysis to generate the visualization atlas set D, where D represents the atlas set containing PRPD, PRPS and dielectric spectrogram; S5: Based on the knowledge graph, perform fault diagnosis on the visualization atlas set D. The knowledge graph uses equipment components as nodes E={e_1,e_2,…,e_m}, and electrical connections and physical adjacencies as edges F. Combine the PRPD and PRPS characteristic values to determine the fault location, type, gas chamber state and decomposition product concentration, and generate the diagnosis result set G, where G represents the structured set containing fault information; S6: Evaluate the insulation state risk based on the diagnosis results, analyze the defect trend, safety and economic risks, calculate the risk value through FMEA and FTA methods, and output the alarm level and treatment suggestions.

2. The insulation condition detection and evaluation method for the GIL equipment of a hydropower station based on an industrial Internet platform according to claim 1, wherein: The sub-steps of S1 are: S1.1: Collect voltage signals, current signals, temperature data, vibration characteristics, UHF signals and ultrasonic signals from the GIL equipment through on-line monitoring, live detection or outage tests, and generate the initial data set A; S1.2: Use preprocessing techniques to denoise and standardize the initial data set A to obtain the data set Q, where denoising is completed through wavelet transform and normalization is performed using the mean normalization method; S1.3: If the outliers of the voltage signal and current signal in the data set Q exceed the preset threshold, calculate the Pearson correlation coefficient between the voltage signal and the current signal through correlation analysis to determine whether there are potential defects in the insulation state; S1.4: According to the obtained data set Q, extract the temperature data and vibration characteristics, and use the principal component analysis algorithm to reduce the dimension of the temperature data and vibration characteristics to obtain the data set R; S1.5: If the variance of the feature vectors of the UHF signal and ultrasonic signal in the data set R is lower than the preset threshold, classify the UHF signal and ultrasonic signal through clustering analysis to determine the category of the insulation state; S1.6: Obtain the Pearson correlation coefficient and category information, and use the support vector machine algorithm to classify and train the data set Q to obtain the prediction model of the insulation state; S1.7: Test the dataset R through the prediction model to obtain the classification result of the insulation state of the GIL device.

3. A method for detecting and evaluating the insulation state of a hydropower station GIL device based on an industrial Internet platform according to claim 1, characterized in that: The sub-steps of S2 are as follows: S2.1: Process the original feature set A through photoelectric conversion, use a photoelectric sensor to obtain the initial signal data, and obtain the converted signal S; S2.2: Filter the converted signal S, use a low-pass filter to remove high-frequency noise, and obtain the filtered signal T; S2.3: Amplify the filtered signal T, use an operational amplifier to enhance the signal strength, and obtain the amplified signal U; S2.4: Perform time-domain analysis on the amplified signal U, calculate the time-series statistics of the signal, and obtain the vibration amplitude feature V; S2.5: Perform frequency-domain analysis on the amplified signal U, extract frequency components through fast Fourier transform, and obtain the UHF signal feature W and the ultrasonic signal feature X; S2.6: If the integrity of the vibration amplitude feature V, the UHF signal feature W, and the ultrasonic signal feature X meets the preset threshold, generate the structured feature set B through the feature fusion algorithm; S2.7: If not, return to the signal filtering step for reprocessing and judge the integrity of the feature set; S2.8: Perform normalization processing on the structured feature set B, use the z-score normalization method, and obtain the normalized dataset Y.

4. A method for detecting and evaluating the insulation status of a hydropower station GIL device based on an industrial Internet platform according to claim 1, characterized in that: The sub-steps of S3 are as follows: S3.1: Obtain the original data from the structured feature set, use a predefined serialization format to generate a JSON data packet, and obtain the serialized dataset; S3.2: Encrypt the JSON data packet through an encryption algorithm to generate an encrypted data packet and determine the transmission security; S3.3: If the integrity verification of the encrypted data packet passes, transmit it to the substation data center through an encrypted channel to obtain the transmission completion status; S3.4: Receive the encrypted data packet from the substation data center, decrypt it to generate the stored feature set, and obtain the stored dataset; S3.5: If the stored feature set is consistent with the serialized dataset, send it to the big data center through the synchronous transmission protocol to determine the synchronous completion status; S3.6: According to the stored feature set received by the big data center, use a data verification algorithm to judge the data integrity; S3.7: If the data integrity verification passes, store the feature set in the big data center to obtain the final storage status.

5. The insulation condition detection and evaluation method of a hydropower station GIL device based on an industrial Internet platform according to claim 1, characterized in that: The sub-steps of S4 are as follows: S4.1: Obtain the feature values from the stored feature set C, calculate the phase-resolved features through partial discharge analysis, and obtain the initial data of the PRPD diagram and the PRPS diagram; S4.2: Process the initial data of the PRPD diagram using Fourier transform, extract the phase-resolved discharge features, and obtain the frequency-domain distribution of the discharge diagram; S4.3: Process the initial data of the PRPS diagram through time-series analysis, determine the time distribution characteristics of the pulse sequence, and obtain the time-domain features of the pulse sequence diagram; S4.4: Extract the dielectric response data from the feature values, use the dielectric spectroscopy analysis algorithm to generate the frequency-domain response distribution of the dielectric spectrogram; S4.5: If the frequency domain distribution of the PRPD diagram matches the preset discharge feature threshold, then combine the time domain features of the PRPS diagram to generate a joint feature set of the discharge pattern; S4.6: Obtain the joint feature set and the frequency domain response distribution of the dielectric spectrogram, and use a data fusion algorithm to obtain a visualization atlas set D that includes the PRPD diagram, the PRPS diagram, and the dielectric spectrogram; S4.7: Through visualization processing, convert the atlas set D into a two-dimensional image for output to obtain the final visualization result.

6. The insulation condition detection and evaluation method for the GIL equipment of a hydropower station based on an industrial Internet platform according to claim 1, wherein: The sub-steps of S5 are: S5.1: By parsing the visualization atlas set D, obtain the device component nodes E = {e_1, e_2,..., e_m} and the electrical connection and physical adjacency edges F from the knowledge graph, construct a topological relationship matrix Z, and obtain the connection features between device components; S5.2: According to the topological relationship matrix Z, fuse the PRPD eigenvalue, and use the k-nearest neighbor algorithm to calculate the abnormal probability P(e_i) of each device component. If P(e_i) is greater than the preset threshold T_1, then determine that the component e_i is a potential fault location to obtain a fault location set α; S5.3: For each component e_i in the fault location set α, extract the corresponding PRPS eigenvalue, use the support vector machine algorithm to classify the fault mode, and determine whether the fault type is breakdown and discharge or overheating to obtain a fault type set β; S5.4: Obtain the chamber state parameter S(e_i) of the component e_i from the fault location set α. If S(e_i) deviates from the preset threshold range T_2, then determine that the chamber state is abnormal to obtain a chamber state set γ; S5.5: According to the fault type set β and the chamber state set γ, analyze the decomposition product characteristics. If the fault type is discharge and the chamber state is abnormal, then calculate the decomposition product concentration C through spectral analysis technology to obtain a decomposition product concentration set δ; S5.6: By fusing the fault location set α, the fault type set β, the chamber state set γ, and the decomposition product concentration set δ, generate a structured fault diagnosis result set G, where G = {α, β, γ, δ}; S5.7: If there are multi-component fault associations in the fault diagnosis result set G, based on the electrical connection edge F of the knowledge graph, use a graph convolutional network to update the fault propagation path to obtain the final diagnosis result set G.

7. A method for detecting and evaluating the insulation status of a hydropower station GIL device based on an industrial Internet platform according to claim 1, characterized in that: The sub-steps of S6 are: S6.1: According to the diagnosis result set G, perform a fault mode and effect analysis, identify each fault mode and its influence range on the system performance and component operation, and combine the edge F of the knowledge graph to generate a component association influence set H, where H represents the mapping set of the fault mode and the affected components; S6.2: According to the component association influence set H, perform a fault tree analysis, define insulation failure as the top event, use the fault modes in H as basic events, construct a causal propagation path, and generate a fault propagation atlas I, where I represents the logical relationship set from component failure to system failure; S6.3: According to the diagnostic result set G and the fault propagation graph I, quantify the defect trend, safety risk, and economic risk of the insulation state, calculate the comprehensive risk value J, where J represents the risk quantification value, generate the alarm level K and the handling suggestions L, where K represents the risk level and L represents the set of suggested measures; S6.4: If the alarm level K is higher than the preset threshold M, where M represents the risk trigger threshold, then perform convolutional neural network analysis on the UHF signals and ultrasonic signals in the structured feature set B. The convolutional kernel is based on a pre-trained model to generate the fault location classification probability N, where N represents the location verification probability, and update the fault location in the diagnostic result set G; S6.5: Based on the fault location in the updated diagnostic result set G, query the associated device components through the knowledge graph. The edge F is defined as the electrical dependency relationship to generate the fault impact range O, where O represents the set of components affected by the fault {e_i, e_j,...}; S6.6: According to the fault impact range O, the diagnostic result set G, and the alarm level K, generate the diagnostic report P based on the predefined alarm rules. P includes the fault type, location, impact range, and handling suggestions, and transmit P to the monitoring system to complete the insulation state monitoring.

8. A method for detecting and evaluating the insulation status of a hydropower station GIL device based on an industrial Internet platform according to claim 1, characterized in that: The specific method of S6.1 is as follows: Obtain the fault mode data from the diagnostic result set, parse and extract the feature information to form a fault mode set; for this set, combine the system performance parameters and the component operating status, and use the impact analysis algorithm to determine the impact range of each fault mode on the system and components; with the help of the knowledge graph, obtain the component association relationship according to the edge F to generate the preliminary association impact set of the fault mode and components; if there are multi-path impacts in the preliminary set, use the depth-first search algorithm to traverse the edge F to judge the propagation path to obtain the optimized association impact set; construct the mapping relationship between the fault mode and the affected components according to the optimized set to form a component mapping set; extract the identification information of the affected components from the component mapping set, and combine the system performance change trend to judge the priority to obtain the final component association impact set H; verify the consistency between H and the diagnostic result set through data comparison, determine the accuracy of H and output the final mapping set.

9. The insulation condition detection and evaluation method for the GIL equipment of a hydropower station based on an industrial Internet platform according to claim 1, characterized in that: The specific method of S6.2 is as follows: Obtain the component association impact set H, and extract the fault modes from it to form the basic event set ε; use the fault tree analysis method, set the insulation failure as the top event ζ, and construct the fault tree structure η based on ε; through Boolean algebra operations, analyze the causal propagation relationship between the basic events in η to generate the causal propagation path θ; calculate the logical relationship strength ι of each path according to θ to obtain the fault propagation graph I; If the ι of a certain path in I exceeds the preset threshold, mark this path as a high-risk path κ; Extract the key basic events λ with the help of κ to generate the optimized fault tree μ; finally, output the system failure logical relationship set ν of the complete causal chain from the key component failure to the insulation failure according to μ.

10. A method for detecting and evaluating the insulation status of a hydropower station GIL device based on an industrial Internet platform according to claim 1, characterized in that: The specific method of S6.3 is as follows: Extract the insulation state data from the diagnostic result set G, and use statistical analysis to obtain the defect characteristics and change trends; Combined with the defect change trend and the fault propagation map I, use the graph analysis algorithm to calculate the defect propagation path and clarify the distribution of potential fault points; If the distribution of potential fault points exceeds the preset threshold, obtain the safety risk value ξ through the safety risk model to complete the quantification of safety risk; use the economic loss assessment model to calculate the economic risk value ο in combination with the defect change trend to achieve the quantification of economic risk; Adopt a weighted fusion method to comprehensively obtain the comprehensive risk value J from the safety risk value ξ and the economic risk value ο, and determine the comprehensive risk quantification index; if the comprehensive risk value J exceeds the preset threshold, generate the alarm level K according to the risk grading rules; Finally, according to the alarm level K, extract the corresponding set of treatment measures L from the preset measure library to generate recommended measures.

Images